Three shifts in course design to update pedagogies for the AI era

Most students succeed in well-structured courses with clear objectives, predictable assessments and fair grading. Yet when they enter professional life, graduates are often tasked with ambiguous problems, incomplete data and issues they’ve never faced before – now under time pressure. 

Generative AI widens this gap. Students can now access vast amounts of knowledge instantly, but easy access does not automatically produce the critical judgement, contextual understanding or adaptive problem-solving that workplaces demand. 

• Spotlight guide: Course design for a new age
• ‘Small changes in assessment design can make thinking visible’
• Your students use GenAI. Let’s redesign learning so thinking still happens

Without deliberate guidance, AI risks becoming a tool for cognitive offloading rather than deeper learning. Updating traditional pedagogies requires three shifts in course design, with the educator moving from primary knowledge transmitter to coach.

Build students’ capacity to work with uncertainty through supported challenge

Traditional course design often seeks to minimise uncertainty, enabling students to demonstrate their mastery more efficiently. While beneficial, this can leave graduates underprepared for professional ambiguity and complexity. AI might provide instant access to information, but higher education’s value lies in developing the judgement to work effectively, even with uncertainty and incomplete information.

Rather than removing all uncertainty, educators can design tasks that contain genuine unknowns, while providing strong scaffolding. This might involve problems with missing information, conflicting constraints or requirements that evolve partway through. Students still receive clear criteria, access to relevant sources and structured checkpoints, but they don’t have the complete roadmap from the outset.

For example, in one of my experiential learning laboratories, students master essential techniques, then investigate a compound’s pharmacological properties through professional knowledge and reasoning rather than a standard protocol. They decide which tests to perform, interpret ambiguous or unexpected results, and justify conclusions. They’re supported by rubrics and consultations, while uncertainty fosters problem-solving. 

The goal is not to increase their stress, but to create supported opportunities for students to practise prioritising what to focus on, seeking clarification, revising their approach and making reasoned decisions under conditions of uncertainty. These are adaptive capacities that highly structured academic success can sometimes leave underdeveloped.

This approach can be applied in various courses by designing tasks that require students to work with incomplete information or evolving constraints, while still providing appropriate scaffolding and feedback.

Teach students to treat AI as a research companion 

Because AI gives students rapid access to knowledge, your role as educator is now coaching how to interrogate, verify and apply knowledge wisely. Without coaching, students bypass synthesis and judgement.

Require an AI interaction log for every major assignment: Students record prompts used, outputs received and – crucially – what they did next: which claims they verified against primary sources, which hallucinations or biases they identified and which suggestions they rejected as unfit for the context. Assess both the log and the quality of their critical commentary, not only the final product.

Design tasks that make over-reliance visible: Students generate initial ideas themselves, then use AI for searching or outlining while remaining responsible for analysis and recommendations. They must identify three specific weaknesses the AI failed to flag and suggest improvements. Early, low-stakes versions of these tasks allow discussion of what effective AI companionship looks like: AI provides speed and breadth; humans must supply depth, contextual judgement and accountability.

The coaching emphasis moves the focus from “what do you know?” to “how do you know, and how did you decide what matters?” Feedback on students’ logs and critical reflections becomes central to learning.

Anchor learning in real-world challenges as the organising principle

Connecting classroom work to authentic problems should shape core module design from the beginning rather than as an add-on.

In my bioindustry course, industry experts are invited to contribute real-world problems; they present challenges like regulatory changes or sustainability issues and provide feedback on student solutions. This exposes students to ambiguous problems and adds authenticity often absent from academic tasks. Students research backgrounds with AI, identify uncertainties, propose responses and present to stakeholders, experiencing the gaps and pressures professionals face.

The pedagogical benefit is direct. Students see why structured thinking is insufficient: real problems lack neat outcomes. They define problems, select evidence and accept residual risk. Educators coach reasoning and justify choices under uncertainty.

Getting started

Select one existing module and apply some of these shifts this semester. Introduce one layer of supported uncertainty into an existing assignment. For courses with projects or major written work, add an AI-interaction log with targeted questions on verification and judgement. You can also bring authentic real-world challenges into the course – for example, by inviting external input or using current case studies. Keep most of the course familiar so the change feels manageable. Gather student feedback on workload and value, then refine and expand next year.

Traditional pedagogies provided reliable transmission and equitable assessment. The 2026 update adds deliberate practice in working with uncertainty, critical AI partnership and real-world problem ownership, with educators guiding students to think, verify and decide amid abundant, but imperfect, information.